COSMOS GALAXIES PLANETS - Prof. Joel Primack Research Projects Physics 205 - 23Feb2021
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Hubble Space Telescope Ultra Deep Field - ACS This picture is beautiful but misleading, since it only shows about 0.5% of the cosmic density. The other 99.5% of the universe is invisible.
Many stars in the very early universe may have been much more massive than our sun, in binary star systems with other massive stars. When these stars ended their lives as supernovas, they became massive black holes or neutron stars. The Laser Interferometer Gravitational-wave Observatory (LIGO) has now detected > 50 mergers of massive black holes. This confirmed key predictions of Einstein’s general relativity for the first time. In August 2017 LIGO and VIRGO announced the discovery of gravity waves from merging neutron stars. Data from telescopes shows that such events probably generate most of the heavy elements like europium, gold, thorium, and uranium.
Matter and Energy Content of the Dark Universe Matter Ships on a ΛCDM Dark Double Energy Dark Ocean Imagine that the entire Theory universe is an ocean of dark energy. On that ocean sail billions of ghostly ships made of dark matter...
CDM Structure Formation: LinearMatter Theory and Energy Content growth by 105 x of the Cold DM inside horizon Universe outside horizon Hot DM Cluster and smaller-scale ν fluctuations damp because of “free-streaming” Scale factor a = 1/(1+z) log a ➞ Cold DM CDM fluctuations that enter the horizon during the radiation dominated era, with masses less than about 1015 M , grow only ∝ log a, because they are not in ⦿ the gravitationally dominant component. But matter fluctuations that enter the horizon in the matter- dominated era grow ∝ a. This explains the characteristic shape of the CDM fluctuation spectrum, with δ(k) ∝ k-n/2-2 log k and δprimordial = kn Primack & Blumenthal 1983, Primack Varenna Lectures 1984 Blumenthal, Faber, Primack, & Rees 1984
CDM Structure Formation CDM Structure Formatio Milky Way mass Dwarf galaxies galaxies start start forming forming Clusters start forming growth by 105 x inside horizon outside horizon t 0.4 20 500 Myr z 1000 100 10 Scale factor a = 1/(1+z) log a ➞
Matter Distribution Agrees with Double Dark Theory! Hlozek et al. 2012
Cosmic Background Cosmic Background Radiation Radiation Angular scale 6000 6000 Planck Collaboration: Cosmological parameters 5000 90 18 1 0.2 0.1 5000 0.07 European European 6000 Double 6000 4000 4000 Planck Collaboration: The Planck mission Planck Collaboration: The Planck mission [µK2 ]2 ] Double Space DDTTTT[µK 5000 3000 Dark Space 3000 5000 Dark Planck Collaboration: Cosmological parameters Theory Theory 4000 2000 Agency 2000 [µK ] ] Agency 4000 [µK 2 3000 1000 1000 PLANCK Cosmic T2T PLANCK 3000 Variance 00 D 2000 600 600 60 60 Satellite 300 300 30 Fig. 7. Maximum posterior CMB intensity map at 5 resolution derived from the 30 DDTTTT Satellite 2000 0 joint baseline analysis of Planck, WMAP, and 00 0 0 408 MHz observations. A small strip of the Galactic plane, 1.6 % of the sky, is filled in by a constrained realization that has the same 1000 D statistical properties as the rest of the sky. -300 -30 Temperature-Temperature 0 Fig. 7. Maximum posterior CMB intensity map at 5 resolution derived from the joint baseline analysis of Planck, WMAP, and 408 MHz observations. A small strip of the Galactic plane, 1.6 % of the sky, is filled in by-30 { -300 Data a constrained realization that has the same Temperature-Temperature -600 statistical properties as the rest of the sky. -60-60 Data 0 -600 600 22 10 10 30 30 500 500 1000 1000 1500 1500 2000 2000 2500 2500 60 300 30 Released 1000 DT T Fig. Fig. 9. Fig.9. The 10.The Planck Planck 0Planck 2015 T T2015 power temperature spectrum. power temperature power The points spectrum. At multipoles in theAtupper spectrum. panel show multipoles ` ` 30the 30 we show wemaximum-likelihood the show the maximum estimates maximum likelihood likelihoodof frequency the primary frequency averaged averaged CMB 0 Released September 24, temperature temperature spectrum-300 spectrum computed computed spectrumascomputed describedfrom from the in thethetext Plik Plik cross-half-mission for cross-half-mission the best-fit foreground likelihood likelihood and nuisance with Fig. with 8. Maximum foreground foreground parameters posterior amplitude Stokes and Q (left) and ofother and U other the Planck+WP+highL (right) maps derived nuisance nuisance parametersfitdeter- from Planck observations parameters between 30 and 353 GHz. deter- listed -30 mined from the MCMC analysis analysis of ofthethebasebase⇤CDM cosmology.InInthe ⇤CDMcosmology. the multipole range 2 ` 29, we plot IX 2015).the power These mapS have been highpass-filtered with a cosine-apodized filter between ` = 20 and 40, and the a 17 % region of the Galactic spectrum February 2019 9, mined from the MCMC in Table 5. The red line shows the best-fit base ⇤CDM spectrum. The lower panel multipole range 2 29, shows the residuals with respect to the theoretical ` we plot the power spectrum plane has been replaced with a constrained Gaussian realization (Planck Collaboration From Planck Collaboration X -60 (2015). estimates -600from the Commander component-separationalgorithm Commandercomponent-separation algorithmcomputed computed over 9494 %% of thethe sky. TheThe best-fit base ⇤CDM theoretical estimates model. from The 0the error bars are computed from the full covariance matrix, over appropriately ofweighted sky. across best-fit each base band (see ⇤CDM Eqs. theoretical 36a and 2015 spectrum spectrum 36b), the and fitted fitted include toPlanck to 2 the Collaboration: the beam Planck TT+lowP 2 Planck TT+lowP 10 uncertainties and likelihoodisisplotted 3010 likelihood uncertainties Cosmological parameters plotted 50 500 in the inthe theupper inforegroundupper 1000 500 These model Fig. panel. 8. panel. Maximum Residuals Residuals beenparameters. consistency posterior inamplitude results between 1000 1500 Stokes theQTT with andand (left) the U full(right) respect viewed as work in progress. Nonetheless, we find a high level of with respect TT+TE+EE One to maps derived 1500 to this 2000 from way of this model Planck observations assessing the model are between constraining power 30 and are contained 2000 shown 2500 353 in shown 8.2.2. Number of modes GHz. a par- in in 2500 the lower lower panel. The error error bars barsshow show±1 uncertainties.From ±1 uncertainties. FromPlanck Planck Collaboration not depend strongly onXIII (2015). Agrees with with Double DoubleDark Theory! likelihoods. mapS have Furthermore, thewith highpass-filtered cosmological parameters filter a cosine-apodized (whichbetween = 20 and 40,ofand ` measurement theanisotropies a 17 % region of the Galactic panel. The Collaboration comparable errors XIII (2015). ticular CMB is to determine the plane has been do replaced ⌧) derived from with a constrained Gaussianthe T E spectra have realization (Planck Collaboration IX 2015). From Planck Collaboration X Agrees Dark Theory! Multipole moment, to the T T -derived parameters, and they are e↵ective number of a modes that have been measured. This `m (2015). consistent to within typically 0.5 or better. is equivalent to estimating 2 times the square of the total S/N Temperature-Polarization in the power spectra, a measure that contains all the available Polarization- 100 viewed as work 16 Double Dark Theory Temperature- Double Dark Theory in progress. Nonetheless, we find a high level of 8.2.2. Number of modes 140 consistency in results between the TT and the full TT+TE+EE likelihoods. Furthermore, the cosmological parameters (which One way of assessing the constraining power contained in a par- 20 Fig. 1. Planck 2018 temperature power spectrum. At multipoles ` 30 we show the frequency-coadded temperature sp ticular measurement of CMB anisotropies is to determine the Temperature-Polarization showing a precise Polarization-Polarization do not depend strongly on ⌧) derived from the T E spectra have Fig. 19. The temperature angular computed from the Plik cross-half-mission power spectrum of the primary Polarization CMB from Planck, likelihood, with foreground and other nuisance measurement Polarization of seven parameters acoustic peaks, that fixed to a best fit80as 15000 comparable errors to the T T -derived parameters, and they are consistent to within typically 0.5 or better. e↵ective number of a`m modes that have been measured. This is equivalent to estimating 2 times the square of the total S/N are well fit by a simple six-parameter ⇤CDM theoretical model (the model the base-⇤CDM 70 plotted is the one labelled [Planck+WP+highL] in Planck Collaboration cosmology. In the multipole range 2 ` 29, we plot the power spectrum estimates from the Com in the power spectra, a measure that contains all the available µK2 ] 10 XVI (2013)). component-separationThe shaded area aroundcomputed algorithm, the best-fit curve overrepresents 86 % ofcosmic variance, the sky. Theincluding the sky cut base-⇤CDM used. The error theoretical bars on individual spectrum best fit points to 60 the DT E [µK2 ] 16 10000 also include cosmic variance. TT,TE,EE+lowE+lensing The horizontal likelihoods axis is logarithmic is plotted in light up to `in blue 50, the and linear upper beyond. panel. The vertical Residuals scale with is respect `(` 1)C to this /2⇡. The measured model are sh 5 0 = + l 0 C EE [10 the lower panel. The error bars show diagonal uncertainties, spectrum shown here is exactly the same as the one shown in Fig. 1 of Planck5000 ±1 including cosmic variance (approximated as Gaussian) Collaboration XVI (2013), but it has been rebinned to show better 40 including uncertainties the low-` region. in the foreground model at ` 30. Note that the vertical scale changes at ` = 30, where the horizon switches from -10 logarithmic to linear. -70 20 Double Dark Theory 0 Double Dark Theory the best-fit -20 temperature data Angularassuming alone, scale -140 tected the base-⇤CDM by Planck estimates fromover T Ethe andentire sky, and E E (the which therefore appro “spectrum-based” con- 0 model, adding 90 the18beam-leakage 1 model 0.2 and fixing the0.1Galactic theboth tains baseline Galactic likelihood we and extragalactic Plik use No objects. thepolarization map-basedin-ap 16 dust amplitudes to the central values of the priors obtained from with the polarization efficiencies 6000 10 formation 100 is provided for the sources ⇣ at⌘ this time.the fixed to Theefficienc ⇣ 4 ⌘ PCCS 8 using the 353-GHz maps. This is clearly a model-dependent pro- EE EE tained from the fits on E E: c100 = 1.021; c143 DT E C EE cedure, but 50000given that we fit over a restricted range of multipoles, 0 di↵ers0 from the ERCSC ⇣ EE ⌘ in its extraction philosophy: more e↵ort EE fit 0 where the T T spectra are measured to cosmic variance, the re- -8 0.966; has-100 and been made on the c 217 completeness = 1.040. The CamSpec likeliho of the catalogue, without re- EE fit -10 scribed in the next section, uses spectrum-based e↵ectiv -4 sulting polarization -16 4000 calibrations are insensitive to small changes ducing notably the reliability of the detected sources, whereas in the underlying cosmological model. 1000 ization efficiency 30 corrections, leaving an 1500 overall tempera µK2] 2 10 30 500 1500 2000 2 10 thepolarization ERCSC was built in the 500 calibration spirit free of 1000 to releasing vary within a reliable 2000 catalog a specified pr In principle, the polarization efficiencies found by fitting the
The Hubble parameter H0 is the expansion rate of the universe today. 2 difficulty for ΛCDM is the Hubble parameter tension: A possibly serious Cosmic Background Radiation plus ΛCDM gives H0 = 67.4±0.4 Ia Supernovae & Cepheids Several kinds of nearby observations give H0 = 73.3±0.8 Lensed Quasar Time Delays Verde Treu Riess 2019 “Early Dark Energy,” a 1.brief Figure period Compilation of ~5% of Hubble extra Constant dark and predictions energy at z ~taken measurements 4000, from could the re- resolve this cent literature and presented or discussed at the meeting. Two independent predictions based on
A brief episode of Early Dark Energy } EDE about ~ 35,000 years after the Big Bang modifies the ΛCDM extrapolation of H0 and avoids the Hubble tension. 1.0 Cosmic Density Fraction 0.8 0.6 Radiation ri/rtot Matter Dark Energy 0.4 0.2 0.0 100 101 102 103 104 105 106 Figure 1. Compilation of Hubble Constant predictions and measurements taken from the re- 1+z cent literature and presented or discussed at the meeting. Two independent predictions based on early-Universe data (Planck Collaboration et al. 2018; Abbott et al. 2018) are shown at the top left (more utilizing other CMB experiments have been presented with similar findings), while the Solid curves represent our ΛCDM+EDE middle panel shows late Universe measurements. The bottom panel shows combinations of the late-Universe measurements and lists the tension with the early-Universe predictions. We stress model, and dashed curves are standard ΛCDM with the Planck parameters. Our that the three variants of the local distance ladder method (SHOES=Cepheids; CCHP=TRGB; MIRAS) share some Ia calibrators and cannot be considered as statistically independent. Like- wise the SBF method is calibrated based on Cepheids or TRGB and thus it cannot be considered as fully independent of the local distance ladder method. Thus the “combining all” value should be taken for illustration only, since its derivation neglects covariance between the data. The N-body simulations show that structure three combinations based on Cepheids, TRGB, Miras are based on statistically independent datasets and therefore the significance of their discrepancy with the early universe prediction is forms earlier than in standard ΛCDM, but correct - even though of course separating the probes gives up some precision. A fair summary is that the di↵erence is more than 4 , less than 6 , while robust to exclusion of any one method, the present-day universe is very similar. team or source. Figure courtesy of Vivien Bonvin. Verde, Treu, Riess 2019 Klypin, Poulin, Prada, Primack, et al. 2020
A brief episode of Early Dark Energy 5 HALO ABUNDANCES about ~ 35,000 years after the Big Bang To study halo mass functions we use simulations with 500h 1 Mpc modifies theboxes and 1000h 1 Mpc ΛCDM extrapolation and mesh of H0 size Ng = 7000. Simulations and avoids with larger the 2h 1 Gpc Hubble boxes have lowertension. mass and force resolutions – not sufficient for analysis of galaxy-mass halo abundances. 1.0 Halos in simulations were identified with the Spherical Over- Cosmic density halofinder BDM (Klypin et al. 2011; Knebe et al. 2011) that Density uses the virial overdensity definition of Bryan & Norman (1998). Fraction The resolution was not sufficient for identifying subhalos, so only 0.8 distinct halos are studied. Figure 10 shows the halo mass function at different redshifts. The EDE model predicts more halos at any redshift, but the dif- 0.6 ference is very small at z = 0: a 10% effect for very massive Radiation ri/rtot 15 1 clusters M ⇡ 10 h M and just 1% for Milky Way-mass ha- Matter 12 1 los with M = 10 h M . These differences hardlyDark make any EDE: 50% more cluster at z ~ 1, 2x more massive galaxies at z ~ 4 Energy 0.4 impact on predicted statistics of galaxies and clusters with obser- Figure 10. Halo mass function at redshifts z = 0 4. Full curves in the bot- vational uncertainties and theoretical inaccuracies being larger than tom panel are for the EDE simulations and dashed curves are for the ⇤CDM differences in halo abundances. simulations. The smaller box and better resolution simulations EDE0.5 and 0.2 The situation is different at larger redshifts: the number of ha- ⇤CDM0.5 are used for masses below M ⇠ 2 ⇥ 1013 h 1 Mpc. At z = 0 halo abundances are 0.0 even more at larger redshifts. For example, the EDE model predicts very similar for the models: EDE predicts ⇠ 10% more of the most mas- 100 101 102 103 10512 1 sive clusters M ⇡ 1015 h 1 M and 1%-2% more of galaxy-size halos with 4 6 almost twice more galaxy-size halos with10M > 3 ⇥ 10 h M 10 1+z M ⇡ 1012 13 h 1 M . The differences in abundances increase substantially with the redshift. Klypin, Poulin, Prada, Primack, et al. 2020 Solid curvesandrepresent 4 The broadening our attenuation of the ΛCDM+EDE BAO feature is exponential, t h computed model, and dashed curves are standard as adopted in our model given in Eq. 2, with a scale ⌃ nl Research Projects: following Crocce & Scoccimarro (2006); Matsubara (2008), i.e. ⌃nl th = h ΛCDM Ø with i 1/2the Planck parameters. Our 1 Plin (k)dk . 3⇡ 2 N-body simulations show that structure Run and analyze high-resolution EDE N-body simulations forms earlier than in standard ΛCDM, but MNRAS 000, 000–000 (0000) Fill high-res simulations with galaxies and clusters the present-day universe is very similar. Compare with observations Klypin, Poulin, Prada, Primack, et al. 2020
Prof. Joel Primack Research Projects COSMOS GALAXIES
Aquarius Simulation Volker Springel Milky Way 100,000 Light Years Milky Way Dark Matter Halo 1,500,000 Light Years 18
Bolshoi Cosmological Simulation Anatoly Klypin & Joel Primack 1 Billion Light Years
1 4 3 2 21
Almost all the stars today are in large galaxies like our Milky Way. Nearby large galaxies are disk galaxies like our galaxy or big balls of stars called elliptical galaxies. But most galaxies in the early universe didn't look anything like our Milky Way. Many of them are pickle-shaped and clumpy. We are just now figuring out how galaxies form and evolve with the help of big ground-based telescopes, and Hubble and other space telescopes that let us see radiation that doesn’t penetrate the atmosphere. Hubble Space Telescope Keck Observatory
“Face Recognition for Galaxies” Deep Learning Identifies High-z Galaxies in a Central Blue Nugget Phase in a Characteristic Mass Range Marc Huertas-Company, Joel Primack, Avishai Dekel, David Koo, Sharon Lapiner, Daniel Ceverino, Raymond Simons, Greg Snyder, et al. ApJ 2018 Cosmological zoom-in simulations model how individual galaxies evolve through the interaction of atomic matter, dark matter, and dark energy Our VELA galaxy simulations agree with HST CANDELS observations that most galaxies start prolate, becoming spheroids or disks after compaction events A deep learning code was trained with VELA galaxy images plus metadata describing whether they are pre-compaction, compaction, or post-compaction The trained deep learning code was able to identify the compaction and post- compaction phases in CANDELized images The trained deep learning code was also able to identify these phases in real HST CANDELS observations, finding that compaction occurred for stellar mass 109.5 -10.3 Msun, as in the simulations James Webb Space Telescope will allow us to do even better
“Face Recognition for Galaxies” Huertas-Company, Pre-BN BN Post-BN Primack, et al. ApJ 2018 VELA High-Res Sunrise Images VELA HST-Res Sunrise Images CANDELS HST Images
Convolutional Neural Net (Deep Learning) Galaxy Evolution Phase Determination: HST vs. JWST Deep learning does much HST = Hubble Space Telescope better with JWST images JWST = James Web Space Telescope
Convolutional Neural Net (Deep Learning) High-Redshift Galaxy Giant Clumps: HST vs. JWST Figure 2: E↵ect of varying feedback on the frequency and properties of HST clumps. The figure shows the same VELA galaxy at z ⇠ 2 simulated with two di↵erent feedback strengths (weak: left column, strong: right column). Rows show HST (top) and JWST imag- ing. More low-mass clumps are ob- JWST served in a weak feedback regime. The NIRCam resolution better captures the di↵erence, especially in central regions. simulations suggest that clumps can live long enough to migrate towards the centers of th host galaxies, eventually merging into the progenitors of today’s bulges (e.g., Mandelk et al., 2017). This scenario is supported by observed radial color gradients of clumps (e. Förster Schreiber et al., 2011; Guo et al., 2018). On the other hand, some other mod and simulations suggest that clumps are self-disrupted by their powerful starburst-induc (a) outflows on a timescale of a few tens of Myr (e.g., Genel et(b) al., 2012; Oklopčić et al., 201 FigureThe two scenarios 4: Example distinguish of (a) clump between detection di↵erent with ML feedback and (b) models SED fitting for and strengths, optical detectedas clum properties clumps are very sensitive in CANDELS. A similarto feedback approach (e.g., will Mandelker be followed et al., in this 2017).extending program, Strong feedback the ten to disrupt analysis clumps of clump rather to properties quickly, (Figures from z > 3. preventing themmyfrom JWST proposal.) acquiring large stellar masses a
HST New observatories will have great new capabilities: JWST - higher res, more light, > λ Roman (WFIRST) - 100x HST area Rubin (LSST) - all S sky transients, co-added depth 27
GALAXIES Research Projects: Run and analyze more high-resolution galaxy simulations Convert them into realistic images, including galaxy substructures Use improved Semi-Analytic Models to predict galaxy populations Compare with images from HST, JWST, Roman Space Telescope and ground-based telescopes including Subaru HSC and Rubin
Prof. Joel Primack Research Projects COSMOS GALAXIES PLANETS
We have now discovered about 4000 planetary systems, mainly using star radial velocities from ground-based telescopes and planet-star transits observed by NASA’s satellites Kepler and TESS.
We used to think that our system is typical, with rocky planets near our star and gas giants farther away. Our system also seems unusually “clean” with relatively little debris and dust. We know that there was a “late great bombardment” of the inner planets about 800 million years after the solar system formed. It seems likely that there was a gigantic rearrangement of the Solar System back then.
There may be galactic habitable zones — not too close to galaxy centers where there are frequent supernovae, nor too far where metals (elements beyond He) may be too rare to form rocky planets. Of the ~ 4000 planetary systems astronomers have discovered, there are very few like ours, with all the planets widely spaced in nearly circular orbits. Most planetary systems are much smaller. The most common type of planet seems to be 2 to 6 times Earth’s mass, a “super-Earth”. No such planet exists in our Solar System. Some planets are in the habitable zone around their stars in which water would be in liquid form, but most of these planets are probably not hospitable to advanced forms of life. For one thing, they might not have an optimal abundance of the long-lived radioactive elements thorium and uranium to power a magnetic dynamo and plate tectonics. Too much Th and U would result in a lava world with frequent flood vulcanism, which caused the greatest mass extinction events on Earth. Our living Earth may be a rare “Goldilocks” planet with just the right amount of Th and U.
Radiogenic Heating and its Influence on Rocky Planet Dynamos and Habitability Francis Nimmo, Joel Primack, S. M. Faber, Enrico Ramirez-Ruiz, and Mohammadtaher Safarzadeh 10 Nimmo et al. Astrophysical Journal Letters (November 2020) 3x Earth’s Th and U Earth’s Th and U ⅓ Earth’s Th and U
3x Earth’s Th and U No magnetic dynamo & frequent flood vulcanism Earth’s Th and U Magnetic dynamo & plate tectonics ⅓ Earth’s Th and U Magnetic dynamo but no plate tectonics 34
PLANETS Research Projects: Predict radioactive heating habitability of rocky exoplanets using their star’s Europium abundance Run 2D and 3D simulations of rocky planets, to verify and improve on our 1D modeling Examine other long-term constraints on habitability of rocky planets: changes in orbits, obliquity, effects of supernovas …
Prof. Joel Primack Research Projects Physics 205 - 23Feb2021 COSMOS GALAXIES PLANETS
Some Concluding Thoughts Without Dark Matter We Wouldn’t Exist With only the ordinary matter, the universe would be a low-density featureless soup Dark matter started to form structures very early Galaxies formed within bound “halos” of dark matter Stars formed within galaxies, and stars made elements beyond hydrogen and helium: carbon, oxygen, … Rocky planets formed from these heavier elements Life began and evolved on one such planet Dark matter is our ancestor and our friend! Science Is Much Stranger Than Fiction Before the discovery that most of the density of the universe is invisible, no one imagined this What else remains to be discovered?
Joel Primack RECENT PhD STUDENTS Rachel Somerville (PhD 1997) Jerusalem (postdoc) – Cambridge (postdoc) – Michigan (Asst. Prof.) – MPI Astronomy Heidelberg (Professor) – STScI/Johns Hopkins – Rutgers (Professor) Michael Gross (PhD 1997) Goddard (postdoc) – UCSC (staff) – NASA Ames (staff) James Bullock (PhD 1999) Ohio State – Harvard (Hubble Fellow) – UC Irvine (Professor, Dean) Ari Maller (PhD 1999) Jerusalem – U Mass Amherst (postdoc) – CityTech CUNY (Assoc. Prof.) Risa Wechsler (PhD 2001) Michigan – Chicago (Hubble Fellow) – Stanford U (Prof., KIPAC Dir.) T. J. Cox (PhD 2004) – Harvard (postdoc, Keck Fellow) – Carnegie Observatories (postdoc) – Data Scientist at Voxer, San Francisco – Data Scientist at Apple, Cupertino Patrik Jonsson (PhD 2004) UCSC (postdoc) – Harvard CfA (staff) – SpaceX Senior Programmer Brandon Allgood (PhD 2005) – Numerate, Inc. (co-founder) Matt Covington (PhD 2008) – analytic understanding of galaxy mergers, semi-analytic models of galaxy formation – U Minn (postdoc) – U Arkansas (Assoc. Prof. of Geology) Greg Novak (PhD 2008) – running and comparing galaxy merger simulations with observations – Princeton (postdoc) – Inst Astrophysique de Paris (postdoc) – Data Scientist at Stitch Fix Christy Pierce (PhD 2009) – AGN in galaxy mergers – Georgia Tech (postdoc) – teaching Rudy Gilmore (PhD 2009) – WIMP properties and annihilation; extragalactic background light and gamma ray absorption – SISSA, Trieste, Italy (postdoc) – Data Scientist at TrueCar, L.A. Alberto Dominguez (PhD 2011) – UCR (postdoc), Clemson (postdoc), Madrid (postdoc) Lauren Porter (PhD 2013) – semi-analytic predictions vs. observations – Data Sci at Facebook Chris Moody – analysis of high-resolution galaxy simulations: galaxy morphology transformations (PhD 2014) – Data Scientist at Square – Chief Data Scientist at Stitch Fix, San Francisco Christoph Lee (PhD 2019) – galaxy simulations vs. observations with AI – Data Sci at Outschool David Reiman (PhD 2020) – astrophysics deep learning applications – AI Scientist at DeepMind Joel Primack CURRENT PhD STUDENTS Viraj Pandya — semi-analytic models compared with observations (with Rachel Somerville) Clayton Strawn — circumgalacticlactic medium: simulations vs. observations James Kakos — combining spectroscopic & photometric redshifts with SORT
You can also read